Regression with generated template parameters.

[emilio]

From 0.29 to 0.31, the following code corresponding to nsBaseHashTable has changed from:

    pub struct nsBaseHashtable_EntryPtr<KeyClass, DataType> {
        pub mEntry: *mut root::nsBaseHashtable_EntryType<KeyClass, DataType>,
        pub mExistingEntry: bool,
        pub _phantom_0: ::std::marker::PhantomData<::std::cell::UnsafeCell<KeyClass>>,
        pub _phantom_1: ::std::marker::PhantomData<::std::cell::UnsafeCell<DataType>>,
    }

To:

pub struct nsBaseHashtable_EntryPtr < KeyClass , DataType > {
    pub mEntry : * mut root :: nsBaseHashtable_EntryType < KeyClass , DataType > < KeyClass , DataType > ,
    pub mExistingEntry : bool ,
    // ....
}

Trying to find a test-case now.

[emilio]

Test-case:

template<class K, class V>
class Entry : public K
{
  V mData;
};

template<typename K, typename V>
class nsBaseHashtable {
  typedef Entry<K, V> EntryType;

  struct EntryPtr {
  private:
    EntryType& mEntry;
    bool mExistingEntry;
  };
};

[emilio]

Bisect says that 61743aa is to blame.

[emilio]

Unfortunately that is not precisely a minimal commit...

[emilio]

I got this one though.